Research map / Label-efficient and robust learning / Recognition and long-tail
Robust Weak Supervision with Variational Auto-Encoders
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use WS-VAE (joint VAE with adaptive weak-label decoder) instead of baselines | weak supervision on Wrench benchmark · 6 Wrench binary datasets | F1-score | Mixed |
| Adding weak-label decoder and VAE joint training (WS-VAE) under accuracy degradation of LFs | YouTube with 50% noisy weak labels · YouTube accuracy-degradation simulation | classification accuracy: about 9% over second best (MV+BERT) | Helped |
| WS-VAE vs baselines under random combined LF degradations | 6 Wrench datasets, 300 conditions · random LF degradation (accuracy, coverage, correlation) | macro F1-score: average improvement 0.057 over best baseline in each experiment; best in 66.3% of conditions | Helped |
| Unsupervised VAE-based joint modelling with coverage-robust inference (no ground-truth labels) | YouTube with varying abstained labels · YouTube coverage degradation | classification accuracy | Helped |
| Adding randomly labelling correlated LF copies (stress test) to WS-VAE | YouTube with appended random LF copies · YouTube correlation degradation | classification accuracy | Helped |
| Advantage of WS-VAE is largest when raw majority vote F1 is low | WS-VAE vs best baseline across conditions · F1 vs raw majority vote quality | macro F1-score | Helped |
Closest papers by meaning
- Universalizing Weak Supervision (ICLR 2022 · similarity 0.90)
- Learning from weak labelers as constraints (ICLR 2025 · similarity 0.90)
- AutoWS-Bench-101: Benchmarking Automated Weak Supervision with 100 Labels (NeurIPS 2022 · similarity 0.90)
- A General Framework for Learning from Weak Supervision (ICML 2024 · similarity 0.88)
- Learning with Partial-Label and Unlabeled Data: A Uniform Treatment for Supervision Redundancy and Insufficiency (ICML 2024 · similarity 0.86)
- Learning from True-False Labels via Multi-modal Prompt Retrieving (ICML 2025 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.